Building Expertise Into Generated Conversations
summary
Large language models produce usable outputs but require deliberate scaffolding: teams use few-shot training to generate utterances, point LLMs at client websites to auto‑build conversational style guides, summarize open‑ended research, and create sample dialogues as raw materials. At runtime they implement an entity‑collection AI node to replace spaghetti logic for reservations, use RAG to ground answers, and accept latency limits for voice. Fully autonomous agents power nonlinear shopping and food‑ordering flows. LLMs remain a stochastic parrot, so label bots as AI (avoid human names) and enforce guardrails and evaluation.
AI-written summary · about
chapters (9)
- 0:00Introduction and context
- 2:24Weird times for conversation design
- 8:09AI in conversational UX design
- 17:43Human conversation complexity
- 25:36Generative AI in design practice
- 31:15Runtime AI applications
- 38:42Risks and guardrails for generative AI
- 42:36Stochastic parrots and counterfeit humans
- 57:06Q&A on AI limitations and UX maturity
related talks
17mBuilding Effective Voice Agents
36mLLMs and AI agents demystified
01h 39mMastering LLMs: From Prompt Engineering to Agentic AI
10mBuilding Conversational AI Agents at Scale
58mA Primer on AI Powered Conversational Experiences and Language Models
18mGenerative AI in Financial Industry
35mRoleplay + AI = How to Personalize AI Chatbots
33mOpening Keynote: LLMs, between fantasy and reality - the poin...